Mobile-phone location data as proxies for workplace presence: usefulness and limits

This study examines the usefulness and limits of mobile-phone location data as proxies for workplace presence. We compare mobility-based population measures with administrative attendance records for public high-school teachers in Osaka Prefecture, Japan, using a prediction-based validation framework. In settings with stable and predictable work schedules, simple hour-of-day models predict workplace presence reasonably well, and mobility-based measures add only limited information. Under geographic extrapolation, they sometimes reduce predictive accuracy. Under time extrapolation, however, their contribution becomes more apparent, especially during the COVID-19 period, when temporal work patterns became less stable. Prediction errors are also larger near major transit hubs and in areas with intense surrounding daytime activity, indicating that workplace presence is harder to isolate from general population flows in such environments. These results suggest that mobile-phone location data are best viewed as a supplementary and conditionally useful source of information, particularly when direct records are unavailable and workplace conditions are changing.